Multicolor 100nm diameter fluorescent beads were imaged to compare standard TIRF to TIRF-SIM and quantify the attainable improvement in lateral resolution (Figure 4A-B). Reconstruction of raw frames into super-resolution images was performed using standard algorithms as outlined in the literature^27,28. It can be seen that TIRF-SIM clearly has significantly higher lateral resolution compared to TIRF. The point spread function (PSF) of a microscope is well approximated by the image of a single sub-diffraction sized fluorescent bead, therefore the PSF and the resolution can be quantified by fitting 2D Gaussian functions to individual beads for each wavelength. The estimated resolution of the microscope based on the mean value of the full width half maximum (FWHM) is 89 nm and 116 nm for 488 and 640 nm TIRF-SIM respectively (Figure 4C). This corresponds to a two-fold improvement in lateral resolution for both wavelengths compared to the theoretical diffraction limited case. Fluorescently labelled amyloid fibrils are also an excellent test sample for demonstrating doubled resolution (Figure 4D). Amyloid fibrils were formed in vitro by incubating β-amyloid labelled with 10% rhodamine derivative dyes (488 nm excitation) for 1 week and subsequently imaging with TIRF-SIM. See reference¹² for more information.

Subcellular structures with high contrast such as emGFP labelled microtubules (Figure 5B, G) or LifeAct-GFP (Figure D) are ideal for TIRF-SIM imaging and yield high contrast super-resolution images. TIRF-SIM imaging using the setup detailed in this protocol enables observation of a sub-population of microtubules located in the vicinity of the basal cell cortex, and microtubule polymerization and depolymerization can be seen over time (Animated Figure 1). Not all samples are amenable to imaging with TIRF-SIM, in particular, low contrast samples without discrete structures. Cells expressing cytosolic GFP lack high resolution information aside from at the edges of the plasma membrane (Figure 5 F, H and Animated Figure 2) and are hence sub-optimal for TIRF-SIM imaging as the resulting reconstructions are essentially TIRF images overlaid with artifacts. In such samples, the increase in contrast can often be attributed to the deconvolution step of the reconstruction algorithm.

High modulation contrast is essential for successful SIM imaging. The Fourier transform of the reconstructed image allows visualization of the SIM optical transfer function (OTF) (Figure 6A, inset). Without maximizing the modulation contrast for each orientation by ensuring azimuthal polarization with a polarization rotator, there is very little modulation of the high-resolution information in the sample leading to a low signal-to-noise ratio in the SIM passbands. Reconstruction algorithms which use the standard Wiener filter approach will simply amplify the noise in the SIM passbands and yield an image which is essentially a standard TIRF image overlaid with hexagonal (or "honeycomb") ringing artifacts (Figure 6A, right panel). A possible enhancement might be the use of iterative^29,30 or blind reconstruction algorithms^31,32 to reduce these artifacts depending on the type of sample. We recommend the use of the ImageJ plugin SIMcheck to check the quality of SIM data before and after reconstruction³³.

Figure 1: Layout of the multicolor TIRF-SIM setup. The TIRF-SIM microscope consists of three main parts, the beam generation unit, the pattern projection unit, and the detection unit. In the beam generation unit, three different lasers are aligned onto the same beam path via dichroic mirrors (DM1 and DM2) and directed through four optical elements for polarization control. First, a polarizer (P) ensures the purity of the linear polarization state of each of the laser beams. The following three optical elements are needed to rotate the polarization in a fast, automated manner as described in detail in the text. Afterwards, two lenses (L1 and L2) in a telescope configuration expand the beam to match the active surface of the spatial light modulator (SLM) and are diffracted into three beamlets by the SLM's projected binary grating patterns (examples are shown in tiles 1-9). The polarization state of the illumination light relative to the SLM pattern is shown as an arrow. A second telescope (L3 and L4) de-magnifies the pattern and offers access to the Fourier plane of the SLM pattern. In this plane a spatial mask (SM) is used to filter out the central component and other unwanted diffraction components from the pixelated structure of the SLM and its internal wiring. Before the two remaining beams are focused onto the back focal plane of the objective (OB) via the tube lens (L5), two dichroic mirrors (DM3 and DM4) are included in the setup. DM4 acts as a conventional dichroic mirror in fluorescence microscopy to separate illumination from emission light. However, this mirror unavoidably induces ellipticity in the polarization state of the illumination light which can be compensated for by DM3, a dichroic mirror from ideally the same batch as DM4. The oil immersion TIRF objective has a large enough NA to directly launch two counter-propagating waves onto the coverslip that are reflected totally and give rise to a structured evanescent field in the coverslip. The sample is mounted on an x-y-z translation stage. Detection is performed through the same objective and DM4 in transmission, plus an additional filtering by bandpass emission filters, mounted in a computer controlled filter wheel (EFW). Finally, the image is projected onto a sCMOS camera by the internal microscope tube lens (L6).

Figure 2: Alignment of overlapping beams. (A)An SLM grating pattern windowed with a circular aperture is useful for alignment. If two non-overlapping beams are visible on the camera (left), then the position of the sample plane must be repositioned by iteratively adjusting the axial positions of the objective lens and the camera to give a single circular illumination spot (right). The beams must overlap in order to produce the sinusoidal excitation pattern required for TIRF-SIM. If the beams do not fully overlap this reduces the field of view over which the interference pattern is formed. (B and C) The precise angle of incidence of the beams is important for TIRF-SIM. If the angle is incorrect, one of the beams will not be at the required angle for TIRF and this is easily visible when imaging a fluorescent dye solution. One beam has an angle of incidence greater than the critical angle which yields the circular spot, and the other does not, which leads to the bright streak on the left of the image in (B). (D) Adjusting the angle of mirror DM3 ensures both beams are incident at the same angle, and this can be validated by defocusing the objective: if correctly aligned, the xz projection of a z stack of a fluorescent dye sample should show two symmetrically intersecting beams with negligible background at the focus.

Figure 3: Synchronization dependencies of the different system components. (A) For fast SIM acquisition, synchronization of the system components using a hardware based solution is essential. (B) A data acquisition board (DAQ) should be used as a master trigger. A TTL signal from the DAQ board is sent to the sCMOS External Input and used to trigger the camera exposure. The camera Global Exposure output then triggers the SLM to display a grating pattern, and the SLM LED Enable output is used to digitally modulate the laser excitation such that the laser is only emitting when the SLM pixels are in the "on" state. After the exposure is complete, the camera Global Exposure output is used to advance the SLM pattern on to the next grating phase or angle. The DAQ board also outputs an analog voltage to the LCVR controller to control the linear polarization state of the illumination beam. This voltage is switched after acquisition of the 3 phase images for each pattern angle. After acquisition of 9 images for a single wavelength, the DAQ board outputs a signal to the emission filter wheel controller, and switches to the next wavelength. The DAQ board also applies a z-offset to the sample by outputting an analog voltage to the z-stage piezo controller.

Figure 4: TIRF-SIM imaging of test samples of 100 nm multicolor beads and fluorescently labelled amyloid fibrils. (A and B) Comparison of standard TIRF compared to TIRF-SIM reconstructions for 488 nm and 640 nm excitation. (C) Histogram of full-width half-maximum (FWHM) of Gaussian fits to the TIRF-SIM beads showing the expected resolution improvement. (D) TIRF versus TIRF-SIM of β-amyloid fibrils labelled with 10% rhodamine derivative dye (488 nm excitation). Scale bars 1 µm.

Figure 5: Live cell TIRF-SIM imaging. Comparison of conventional TIRF and TIRF-SIM images of (A, B) microtubules (emGFP-tubulin) in a HEK293 cell, (C, D) filamentous actin (LifeAct-GFP) in a COS-7 cell and (E, F) cytosolic GFP in a HEK293 cell. Images in B and F are single time points from the movies. Boxed areas are shown magnified in (G, H). Scale bars 3 µm.

Figure 6: Influence of polarization rotator on reconstructed bead images. (A) Without the use of a polarization rotator such as an LCVR, the signal-to-noise ratio in the SIM passbands is low which results in characteristic hexagonal artifacts in the reconstructed SIM images (right), (B) In 2D-SIM, the structured illumination patterns are directly visible in the Fourier transform of the raw images (left, excitation spatial frequency highlighted) as they fall within the radius of the emission OTF support, however in TIRF-SIM, they are outside the OTF support and therefore not visible (right). In this case, the pattern modulation contrast must be assessed using a sparse bead monolayer, as outlined in the protocol.

Figure 7: Spatial light modulator based pattern generation allows implementation of other imaging modalities such as multifocal SIM. (A) In MSIM, a lattice of square pointsdisplayed on the SLM (inset) yields a lattice of diffraction limited foci at the image plane. A thin layer of low concentration rhodamine 6G is imaged to visualize the foci. The pattern is translated across the sample (B) and the acquired raw image z-stack is reconstructed to generate an image with reduced out-of-focus light (C). Scale bars 5 µm.

Animated Figure 1: Time series movie of emGFP-tubulin in a HEK293 cell. Rapid polymerization and depolymerization of emGFP labelled microtubules can be observed using TIRF-SIM. Images acquired using 50 ms exposure time per raw frame (450 ms per SIM frame) spaced at intervals of 0.5 s. Exposure time used was limited by the brightness of the fluorophore, not by the speed of the camera or SLM.

Animated Figure 2: Time series movie of cytosolic GFP in a HEK293 cell. Samples with low contrast such as this are not ideal samples for TIRF-SIM imaging. Retrograde membrane flow can be seen in the TIRF images but TIRF-SIM does not provide any additional information apart from at the cell edges. TIRF-SIM images were acquired using 50 ms exposure time per raw frame (450 ms per SIM frame) spaced at intervals of 5 s.

Supplemental Code File: Example SLM repertoire file (48449_300us_1-bit_Balanced.seq3).

Supplemental Code File: Example SLM repertoire file (period9_001.bmp).

Supplemental Code File: Example SLM repertoire file (period9_002.bmp).

Supplemental Code File: Example SLM repertoire file (period9_003.bmp).

Supplemental Code File: Example SLM repertoire file (period9_004.bmp).

Supplemental Code File: Example SLM repertoire file (period9_005.bmp).

Supplemental Code File: Example SLM repertoire file (period9_006.bmp).

Supplemental Code File: Example SLM repertoire file (period9_007.bmp).

Supplemental Code File: Example SLM repertoire file (period9_008.bmp).

Supplemental Code File: Example SLM repertoire file (period9_009.bmp).

Supplemental Code File: Example SLM repertoire file (period9_mask_1.bmp).

Supplemental Code File: Example SLM repertoire file (period9_mask_2.bmp).

Supplemental Code File: Example SLM repertoire file (period9_mask_3.bmp).

Supplemental Code File: Example SLM repertoire file (TIRF-SIM_example.rep).

Supplemental Code File: Example grating generation code (1 of 2) (generate_gratings.m).

Supplemental Code File: Example grating generation code (2 of 2) (circular_mask.m).

Supplemental Code File: Example code to calculate modulation contrast (calculate_contrast.m).